区块链技术具有分布式、高效和可验证等特点。基于区块链的链下存储系统可以在确保数据完整性和安全性的前提下实现数据的分布式存储。在智慧医疗、物联网和数字城市等分布式数据安全存储至关重要的场景中，具有巨大的应用价值。 本文基于区块链技术，研究快速读取、快速写入和安全优化的链下存储系统。重点解决了链下存储系统在查询过程中快速生成实时不存在证明、大量写入交易下实现高吞吐以及在不同分布式节点上部署恶意流量识别系统三个技术难题，从而实现高效性和安全性的优化，具体内容如下： 1. 提出了加速不存在查询与证明的可验证布隆过滤器。 在传统区块链中存在一个问题，即在过滤“无效查询请求”时没有提供“不存在证明”，这使得恶意节点可以对指定用户发动拒绝服务攻击。为了解决这一问题，本文提出了一种可验证布隆过滤器的构建方式，它能够快速过滤无效查询请求，同时有效地提供证据来证明数据不存在。此外，针对证明过程可能造成的隐私泄露问题，本文还提出了两种解决方案：一是“隐秘的可验证布隆过滤器”，确保每个“不存在证明”只会泄露布隆过滤器的一个置零位，从而减少数据泄露量；二是“数据混淆”，进一步降低用户从泄露的布隆过滤器中推测出真实内容的准确率。 2. 提出了针对写入交易优化的去签名打包机制。 区块链在共识过程中，所有共识节点需要验证每一笔交易的签名，大量的验签操作严重限制了写入交易的吞吐。针对该问题，本文将写入交易去签名后构造默克尔树再提交参与共识，降低共识过程中的验签数量。同时，针对去签名后存储节点可能做出的恶意行为，本文设计了基于智能合约的挑战和应答机制对其进行限制，以同时保证写入交易的高吞吐和安全可认证。 3.提出了基于神经元激活状态的入侵检测系统跨节点迁移技术。利用网络入侵检测系统识别大量访问中的恶意请求是存储系统安全性的重要保障。由于链下存储系统建立在分布式节点之上，不同节点面对的服务流量分布存在差异。传统统一数据集训练的网络入侵检测模型无法适应不同的节点，导致在其他节点上部署会出现大量误报。因此，本文采用神经元激活状态分析方法自动识别并过滤虚假警报，提高网络入侵检测系统在分布式节点下部署的可用性，保障链下存储系统稳定安全运行。

Blockchain technology is distributed, efficient and verifiable. The off-chain storage system based on blockchain can realize distributed storage of data while ensuring data integrity and security. It has huge application value in key scenarios of diversified data secure storage such as smart medical care, the Internet of Things, and digital cities.This research is based on blockchain technology and aims to study and innovate a blockchain storage system with fast read, fast write, and security optimization. The focus is on solving three technical challenges in storage systems: fast generation of real-time non-existent proofs, fast throughput under a large number of write transactions, and security optimization of traffic identification systems on different blockchain nodes. The specific content is as follows:1. Proposed a verifiable Bloom filter for accelerating non-existent queries and proofs.In traditional blockchain storage systems, there is a problem of not providing "non-existent proofs" when filtering "invalid query requests", which allows malicious nodes to launch denial of service attacks against specified users. To solve this problem, this paper proposes a method for constructing a verifiable Bloom filter, which can quickly filter out invalid query requests while effectively providing evidence that the data does not exist. In addition, to address the privacy leakage issues that may arise during the proof process, this paper also proposes two approaches: "secret verifiable Bloom filter" to ensure that each "non-existent proof" only leaks one zero bit of the Bloom filter and "data obfuscation" to reduce the accuracy of users further inferring the actual content from the leaked Bloom filter.2. Proposed a de-signature and packaging mechanism for write transactions.In the consensus process of blockchain, all consensus nodes need to verify the signature of each transaction, and a large number of signature verification operations severely limit the throughput of written transactions. This article will remove the transaction‘s signature and repackage transactions through the Merkle tree before participating in the consensus to reduce the signature verification number. At the same time, nodes may behave maliciously. This paper designs a challenge-and-response mechanism based on smart contracts to restrict them, ensuring both high throughput and secure verification of written transactions.3. Proposed a migration technique for node-to-node traffic identification systems based on meta-learning.Identifying malicious requests in a large number of service accesses is an essential guarantee for the security of the storage system. Since the blockchain storage system is built on distributed nodes, the distribution of service traffic varies among different nodes. Traditional flow identification models trained on a unified dataset cannot adapt to the situations of different nodes, resulting in many false positives on different nodes. Therefore, this paper uses meta-learning to train identification models while training multiple classification models and combining the state of the classification models to determine false positives, effectively reducing false positives and improving the security of the blockchain storage system.